Seed dispersal and seedling recruitment

Presented by Elaine Wright.

Dataset

Typical dataset:

• Seedling densities in quadrats at mapped locations
• A map of all potential parent trees (and their DBH) in the vicinity (how large a vicinity will matter a great deal...) Examples: Ribbens et al. 1993, LePage et al. 2000

New Zealand dataset:

• Seedling densities by substrate type, in quadrats within 7 mapped stands
• Sampling strategy random and adaptive
• Light measured above each quadrat
• Understory cover within and above the quadrat
• A map of all potential parents: includes dbh (<= 10 cm for canopy trees, >= 2.5 cm dbh for subcanopy species

Recruitment submodel

The submodel predicts the spatial distribution and density of seedlings as a function of the distribution and size of parent trees

In SORTIE/BC:

• Fit separate equations for seedling dispersion in range of canopy conditions
• Incorporated seedbed effects in the equations
• Determined the effects of management on seedbed dynamics

New Zealand additions to recruitment

• C1: magnitude of wind direction
• Theta-P: direction of max. dispersal
• Bath term: Expected no. seedlings/m2 in the absence of local parents
• Site: Species-specific differences in seedling density
• Light (optimal, low, high)
• Substrate favourability: Total no. seedlings produced on that substrate by a 30 cm DBH tree summed over the seedling shadow

Issues for New Zealand recruitment

• Podocarps dioecious
• Alpha small for some species - doesn't fit well
• Anisotropy of wind direction
• Understory light
• Epiphytes are potential parents
• Masting behavior - does predation negate the effects?
• Seed and seedling predation

Estimating fecundity and modeling long-distance dispersal

Presented by Mike Papaik.

Objectives

1. Determine the importance of favorable seed bed sites to the persistence of small seeded species such as yellow birch and hemlock

2. Assess the importance of

• spatially explicit seed shadows for forest community dynamics
• impact of long-distance dispersal (LDD) in competitive environment

Models

Interspecific variation in seed fecundity and seedling establishment

• seed mass fecundity model (from Greene and Johnson 1994, 1998)
• Seed mass seed germination and seedling establishment model (ditto)
• Seedbed distribution model (new)

'Bath' seed rain models

• Density-dependent - Local dispersal (new)
• Density-independent - LDD (new)

STR can be described as a function of seed mass for density-dependent

Seedbed substrates

Seed size and LDD

Small seed size => improved colonization potential; large seed size => improved competitive ability in saturated habitats. Seed mass size is at best a very weak negative correlate with LDD into uncompetitive environments.

Long distance dispersal

We use a seed mass weighted 'Bath'-LDD model, which modifies basic lottery model in that the probability of establishment depends on the number of propagules. Lottery models were developed because environmental variability (e.g., disturbance) can reverse competitive interactions and slow down competitive exclusion or even change trends.

Discussion of this talk

Question: Are mast-based fecundities based on resource limitation? No, entirely empirical. Seeds were collected for at least three years to establish fecundity - but there are improvements needed because some species mast on longer cycles than that. The data fit was pretty good though.

Question: Would this be a good theoretical approach if we didn't have actual data? The current parameters were tuned to correctly predict sapling densities, not seedling. This represents a disconnect in SORTIE thinking - we don't capture the seedling-sapling transition - so it's a simplification.

Question: How sensitive are the results to initial conditions? Tests indicate results are sensitive almost exclusively to seed mass.

You'd have to link masting behavior to seed mass. Currently we don't have the needed data.

Seedling dispersion patterns and production across a landscape fertility gradient in northern lower Michigan.

Presented by Rich Kobe.

Hypotheses

1. Seedling production increases with soil resources
2. Seedling production increases w/ species site dominance

• competitive advantage
• pollen limitation
• mycorrhizal feedbacks
• differential seed predation

Field methods

It is difficult to attribute offspring to a particular parent. Inverse estimation method was recently developed that made this problem tractable. Essentially perform a non-linear regression using MLE broken down to two components: offspring production and dispersion. What is powerful about this approach is that seedling recruitment can be broken down to two processes: offspring production and dispersal, enabling comparisons standardized to the number of trees and tree size across sites and years.

Used non-spatial estimates of seedling production - fit to data was good - maybe the spatial component is not that important?

The data is for germinated seedlings - so there could be effects included that relate to germination instead of seed production. But data is being collected for seeds themselves - and it looks like a fairly constant transition between seeds and seedlings, at first look.

Soil resources

Mineral nutrients might be influencing frequency and magnitude of mast cycles. Soil N and exchangeable K have a correlation with seed production for red maple, and K for red oak. In the horticultural literature - huge correlation between K uptake and fruit production.

Black oak, white oak, American beech, white ash do not show relationship of fecundity with soil nutrients but show correlation of fecundity with site dominance. Sugar maple and black cherry show no relationship for either soil nutrients or site dominance.

Discussion on this talk

Question: Site dominance effects - due to seed predator satiation?

Question: Is there a negative feedback relationship with dominance? No negative effects. However, increase in density of adults also increases disease - so might cancel out benefits of increased fecundity across life stages.

Inverse modeling is risky - you need to understand the underlying mechanism. So you need to independently validate - genetically mapping seeds to parents or some such.

Small mammal effects on forest recruitment

Presented by Wendy Ruscoe.

In New Zealand, small mammals were recently introduced - so we don't know the long-term effects on adult trees.
Exotic predators:

• House mice (Mus musculus)
• Black rats (Rattus rattus)
• Polynesian rats (R. exulans)

respond to beech (Nothofagus spp) and rimu (Dacrydium sp) seedfall.

Nothofagus seeding

• Seed falls in autumn
• Seed lays on the ground over winter (6 months)
• Most germination occurs in spring.
• Rodents are incredibly opportunistic and will breed over the winter due to nutritional food (seed)

Seed traps collect seed rain on each plot (we get an estimation of seed crop size).

How many seeds can the mice eat?

How does the effect of an exotic mammal fit predator satiation theory??

Numerical response - how many mice are there?

Functional response - how much seed can one mouse eat given what is available?

Simulation

• Seed rain input
• Mouse seed consumption according to functional response
• Mouse population size changes according to numerical response
• Weekly timesteps

How many seeds survive in the spring?

Functional response - a mouse will eat a thousand seeds a night if food isn't limited.

Numerical response - mouse populations limit themselves eventually but don't know the mechanism.

Modeled result - mice cannot eat all the seed. However, native animal seed eating hasn't yet been taken into account.

Conclusion

We can predict rodent increases so we should be able to predict seed losses due to predation - need to be able to model this in SORTIE.

As unnatural processes are affecting survival to sapling size - we need to know if management intervention is necessary.

Discussion of this talk

When beech is masting and causing mouse populations to explode - this could be causing big problems for non-masting species.

Masting of different species is caused by different triggers - non-synchronous - so you'll have to model each species separately.

Question: Are there any mouse-free locations? Only a couple off-shore islands that have been cleaned off - mice are essentially everywhere.

Question: Is this spatially variable? Will different part of the plots mast separately and how would that affect results? Mice populations explode so rapidly that they would swamp all locations.

Ferns as filters for regeneration

Presented by David Coomes

Ferns are filters for regeneration. They let some species through and not others - so they could have a large effect forest dynamics. In New Zealand, ferns might also be blocks - nothing gets through.

New Zealand site

Ferns are omnipresent in NZ. Why? (a) It's wet (= functional fern sex), and (b) it's an evergreen forest (= no spring window to favour angiosperm spring ephemerals).

There is a fertility gradient in P across New Zealand terraces:

• Alluvial forests - fertile, high P
• Terrace forests - infertile, low P
• Terrace shrubland - like terrace forests but poorly drained

We are interested primarily in the largest fern species.
The structurally important ferns:

• Cyathea smithii - up to 14 m mean 3.7 m
• Dicksonia squarrosa - up to 8 m mean 57 m
• Blechnum discolor - up to 2 m
• Blechnum procerum - up to 50 cm
• Gleichenia dicarpa - up to 50 cm

Are ferns acting as filters keeping conifers out in alluvial forests?

Light

As you go down the fertility gradient, the largest ferns in each environment let more and more light through.

In alluvial forests, it is extremely shaded. Gap vs. understory doesn't matter - ferns block everything. But, on alluvial growing surfaces > 80 cm from ground, many more seedlings get through. Unfortunately, little forest floor is raised.

More saplings exist in terrace forest gaps than in alluvial gaps. Most saplings in alluvials are growing on raised logs.

Proposal

The existence of fewer small trees in alluvial forests is due to ferns. This could be tested with SORTIE.

Small ferns can be treated as substrate, but tree ferns cannot.

Tree ferns host angiosperm seedlings on their trunks - they are important sites for regeneration of certain species which eventually overtop the tree ferns. This seems to be the main substrate for regeneration of these species. So - we have to figure out how to model them. Give them a Z value?
Question: Could you simplify and say that they occupy space near the tree fern? Unknown. It might be oversimplification, so probably not.

Discussion of this talk

Question: Don't both of the tree ferns show a disturbance response (increasing in response to disturbance)? In some sites there are dense stands of tree ferns and in others there isn't. Are they blocking angiosperms or acting successionally?

Question: How are trunk angiosperms getting nutrients? Litter falls on fern tops and nutrients leach down.

General recruitment discussion

Recap - we are identifying additional factors that need to be modeled.

Question: Wind - is it possible now to show dispersal showing different dispersal distances in different directions? It doesn't do it currently but could be done.

Question: How to take into account the effects of the ferns blocking the light vs. acting as an unfavorable substrate? We're going to have to carefully figure out light effects - this question is yet to be answered.

Question: In Quebec - understory vegetation - shrubs could be treated as trees. In gaps - sometimes these shrubs will block light and affect regeneration. This is a challenge - how to model the effects of the understory on light? Different groups will probably need different approaches since the dynamics are so different.

Question: Masting years - how to simulate the frequency and total seed production? Current literature is very general for N. America.

Question: In a model run - there are possibilities for free parameters - this could be used to simulate masting. So we'll generate behaviors that, for instance, evaluate STR as a probabilistic function.

Question: Masting is critical - need some stochasticity in recruitment. Think about recruitment in 3 possible models:

• Purely theoretical - based on David Greene's work
• How we do it now
• Use the wealth of literature on seedbed quality combined with theoretical model in the absence of local empirical model

Question: There is a criticism of SORTIE - data requirements are too large. So how do we leverage existing datasets (general or non-local)? Would that improve overall utility of SORTIE? This is a big issue and should get its own discussion. SORTIE is desirable because of its specificity, but would generalizing some functions and mixing them with site-specific make it more useful?

Question: Understory trees in boreal forests produce less seed than same-size trees in more light. How to model this? We model the mean effect by making fecundity a function of DBH. How important is this effect? Does it over-complicate the model? You won't know until you test it - in general doesn't matter but in some cases could be very important.